Hacker News Reader

Comment by mojzu | Hacker News Reader

Although I’m definitely not convinced enough to make the jump yet, those emulation benchmarks are impressive (and seem to also apply somewhat to general use based on another review I read). Running x86 code faster than native equivalents, using an ARM chip, that will probably be consuming significantly less power doing it is something I did not expect to see in gen 1 of this switch.

View on HN

What caught me off guard was the possibility through Codeweavers to run even 32bit windows software

https://www.codeweavers.com/blog/jwhite/2020/11/18/okay-im-o...

That is impressive, makes me wonder whether there's a potential future where Apple puts chips like these into servers to run existing software faster/on lower power

> a potential future where Apple puts chips like these into servers

Not to put too fine a point on it, but that potential future is already here. Mac mini is a server and off-the-shelf rack mounts are easy to find. [0]

[0] For example: https://www.sonnettech.com/product/rackmacmini.html

Without 10gig ethernet its really limited in that role.

It's interesting you say "not convinced", curious what you are skeptical of? I'm not upgrading because I just bought a 16" Mbpro 2 months ago, but I'm more than convinced this new architecture is the real deal. Unfortunately for me my 6 year old MacBook just died and I have to have something for work.

I reckon it's worth waiting a few months for longer term testing, and probably a little bit longer for software support to catch up. There's also a decent chance that M2 chips are somewhere on the horizon and those might help answer the question of whether these SOCs can scale to enthusiast/professional desktops.

It's probably not fair for me to say I'm not convinced yet anyway, as I'm still developing on a Linux desktop and got rid of my laptop during lockdowns as there wasn't a need for it. If that need comes back though, these laptops look like they could be some really nice options (which would be a first for ARM laptops).

Usually gen 2 hardware is the ideal time to buy. It might be a few years before we get a gen 2 MBP, so imo no worries.

Not OP but Apple Gen1 products have a horrible track record.

Although I’m definitely not convinced enough to make the jump yet, those emulation benchmarks…

Remember, it's not emulation; Rosetta translates the x86 binary into ARM code and runs it. That's why there's a small pause when launching x86 apps—it translates the app on the fly.

Transpiling is not the same as emulation.

Someone correct me if I'm wrong, but I think this is mainly because Apple does mostly translation, while Windows on Arm is pure emulation.

This combined with the fact that Arm-based Windows PCs have mostly used mobile chips not optimized for the desktop use-case and form factor, is why the new Arm-based Macbooks leave any Arm-based Windows PCs in the dust in terms of performance.

I’ve heard Windows emulation is also cached JIT.

Also, Apple has tuned the M and A series chips to perform well with Objective-c and Swift as used by iOS and Mac OS. One example is that they made some object operations 5x faster than on Intel chips. This speeds up loops and garbage collection significantly. Those are very common operations.

Other vendors are using generic chips that were not tuned for the OS.

MacOS / iOS apps are not garbage collected

Reference counting is a form of garbage collection, although not usually refered to that way.

Rosetta 2 probably also have a JIT mode to deal with software that JIT produces x86 instructions like browsers and java VM. And of course you can do AOT or trans compile static x86 executables. I feel a lot of success of rosetta comes from the sheer perf of the m1 so even the jit and transcompile process is fast, and the execution of not well optimized code is fast. The firestorm cores is at least 60% faster than snapdragons cores per annandtech

Windows on ARM also runs x86 apps via translation, but in their marketing materials they call it emulation.

x86 emulation works by compiling blocks of x86 instructions into ARM64 instructions with optimizations to improve performance. A service caches these translated blocks of code to reduce the overhead of instruction translation and allow for optimization when the code runs again.

https://docs.microsoft.com/en-us/windows/uwp/porting/apps-on...

Reply on news.ycombinator.com